(UK English Speaking Only)
We’re looking for a Python and Machine Learning expert to build the core visualization engine for an AI-powered floor replacement tool—think along the lines of HomeDesigns.ai Floor Editor or Winstars.ai Flooring Designer. This gig is all about the backend visualization: take an uploaded room image, detect the floor, and swap it with a custom texture, keeping it lean and cost-effective on a server. No fancy frontend—just a simple Python UI for testing.
**Here’s what we need:
Core Task: Use Python and OpenCV to detect floors and replace them with a custom tiled image we provide.
AI Models: Leverage or fine-tune lightweight ML models (e.g., segmentation) for accurate floor detection—good enough to keep things clean and precise. TensorFlow or PyTorch are our go-tos.
Server Efficiency: Deploy it on a low-cost server (like a small AWS/GCP instance). We want a solid balance of speed and accuracy without racking up big server bills—efficiency is the name of the game.
Simple UI: Build a basic Python UI (e.g., Tkinter or a script) so we can upload an image, select a texture, and tweak it—specifically, resize and rotate the tile/texture to fit the room. Nothing flashy, just functional.
Success Looks Like: The final image has no (or barely any) artifacting—clean edges, no weird glitches. The texture’s perspective matches the room realistically, like it belongs there. Plus, we can adjust the tile/texture size and rotation for a perfect fit.
**What We’re Looking For:
Strong Python and OpenCV skills—floor detection and texture replacement should be second nature to you.
Experience with lightweight ML models (e.g., U-Net, MobileNet) that deliver accuracy without eating resources.
Know-how to deploy on a server and keep costs low—think CPU-friendly or minimal GPU, maybe some model optimization tricks like quantization.
Ability to whip up a simple Python GUI that lets us resize and rotate textures on the fly.
A practical approach to hit our success criteria—clean output, realistic perspective, and flexible textures—without overcomplicating things.
**Project Scope & Budget:
Fixed-price job, aiming for a working prototype in 4 weeks.
This is just the visualization engine—no frontend needed.
**How to Apply:
Examples of Python/OpenCV or ML projects you’ve done (code, demos, or write-ups are awesome).
A quick rundown of how you’d tackle this—model ideas, optimization approach, etc.
Your availability and rate. |
